Polarization quanta in Berry phase calculations

[huantd]

Dear VASP community

I am trying to calculate the polarization of a polar solid phase by either LCALCPOL or LBERRY and it turns out that some polarization quanta (eRa/V, eRb/V, or eRc/V with Ra, Rb, and Rc are 3 cell vectors) must be subtracted from or added to the obtained results. In practical cases, it is extremely hard to determine how many quanta we need to add or subtract. Could you please give us some hints?

Regards
Huan

[Neutrino]

Dear Huan,

Usually one needs a centrosymmetric phase to help in removing some of the ambiguities of the polarization quantum. But in spite of having a centrosymmetric phase, still some issues might persist. A good paper that can help in grasping some of these issues is the following:

http://journals.aps.org/prb/abstract/10 ... .71.014113

Mostafa

[huantd]

Dear Mostafa, Admins, and Community

Thanks for the reply and the very nice paper. Yes, an as good as possible centrosymmetric structure is needed, but even in this case, the polarization quantum problem remains unsolved. In small structures with several atoms, the polarization quanta are large enough (~50 muC/cm^2) for some intuitive selection of how many quanta but if the structure size is larger, the quanta become smaller very rapidly. As the quanta approach ~ 5muC/cm^2 (20-30 atoms), it is impossible to know how many quanta are needed. Do you have any systematic strategy/procedure for pulling our such the information from the Berry phase calculation runs?

Thanks
Huan